A photovoltaic array fault detection method based on IV characteristics and data model

Abstract

The present invention discloses a photovoltaic array fault detection method based on IV characteristics and data models, which relates to the field of fault detection technology. The method first determines the acquisition frequency Cp by comprehensively considering dynamic characteristics, historical faults and occurrence speed to ensure that key fault information is not missed, and then relies on the second-order derivative analysis and squareness calculation of the IV curve, combined with the data model, to accurately identify local shadow faults, changing the previous fault judgment that relies on a single indicator. In terms of energy conversion efficiency optimization, the local maximum power point is determined based on the judgment factor Pz and the IV curve, and the photovoltaic array output is dynamically matched with the inverter input. Compared with the traditional fixed matching method, the power generation efficiency is further improved and energy waste is reduced. By starting the model verification mechanism, the accuracy and stability of the model can be verified in different scenarios and time spans.

Description

Technical Field

[0001] The present invention relates to the technical field of fault detection, and in particular to a photovoltaic array fault detection method based on IV characteristics and a data model. Background Art

[0002] In the field of renewable energy, photovoltaic power generation, as an important way to utilize clean energy, is receiving more and more attention. Research based on IV characteristics and data models provides a key way to gain a deeper understanding of the operating mechanism and performance evaluation of photovoltaic arrays. In this specific field, the precise setting of the acquisition frequency and the effective detection of photovoltaic array faults have become the core elements to ensure the stable and efficient operation of photovoltaic systems. By collecting and analyzing the relevant data of the photovoltaic array, it is possible to obtain the IV curve reflecting its working status, and then use the data model to mine potential information, laying the foundation for subsequent fault diagnosis and performance optimization.

[0003] However, there are some shortcomings in the current acquisition frequency setting and photovoltaic array fault detection. In the acquisition frequency determination link, most methods consider a single factor in isolation, which may cause the acquisition frequency to be out of touch with actual needs. In the fault detection link, traditional methods are difficult to accurately identify complex faults such as local shadows. For example, when analyzing the IV curve, it is difficult to effectively use the curve characteristics to accurately judge the fault, and misjudgment or missed judgment often occurs, making it difficult to detect and handle photovoltaic array faults in a timely manner. This will make it difficult for the inverter to make timely adjustments based on the output of the photovoltaic array, seriously affecting the normal operation and power generation efficiency of the system, and hindering the sustainable development of the photovoltaic power generation industry. Summary of the invention

[0004] In view of the above problems existing in the prior art, the present application provides a photovoltaic array fault detection method based on IV characteristics and a data model.

[0005] The embodiment of the present disclosure provides a photovoltaic array fault detection method based on IV characteristics and a data model, comprising the following steps:

[0006] S1. Using several groups of monitoring instruments in advance, real-time monitoring of relevant condition data of the photovoltaic array is performed within a set monitoring period, and a monitoring group is generated after data processing;

[0007] S2. Based on the monitoring group, analyze the impact of the dynamic characteristics of the photovoltaic array on the acquisition frequency, and analyze the impact of the occurrence speed of each fault type on the acquisition frequency in combination with historical data, and determine the acquisition frequency Cp by the maximum value method;

[0008] S3. Based on the acquisition frequency Cp value, data is collected from the photovoltaic array. After the data is collected, the IV curve is drawn. According to the IV curve, the forward difference method is used to estimate the second-order derivative of the discrete data points in the IV curve. , based on the second-order derivative The state changes of the local shadow fault are preliminarily marked, and the second-order derivative is traversed The sequence is statistically preliminarily marked, and the squareness F is obtained based on the degree to which the IV curve tends to be rectangular. Combined with the trained data model, the output judgment factor Pz is fitted to determine the possibility of local shadow failure in the photovoltaic array;

[0009] S4, based on the determination factor Pz and the IV curve, determining multiple groups of local maximum power points to obtain IV characteristic data of each local maximum power point, and dynamically matching the output of the photovoltaic array with the input of the inverter based on the IV characteristic data of each local maximum power point;

[0010] S5. Start the model verification mechanism to verify the credibility of the model.

[0011] Optionally, the specific steps of S1 include:

[0012] S11, deploying several groups of monitoring instruments around the photovoltaic array in advance, and monitoring relevant condition data of the photovoltaic array in real time during the set monitoring period, wherein the relevant condition data includes the light intensity Gq and temperature Wz at each monitoring moment; the several groups of monitoring instruments include photodiode sensors, current sensors, voltage sensors and temperature sensors;

[0013] S12. Check whether the relevant condition data collected in S11 have abnormal values ​​and missing values, and use statistical methods to correct the abnormal values ​​and use interpolation methods to supplement the missing values. Finally, the processed relevant condition data is used as the monitoring group.

[0014] Optionally, the specific steps of S2 include:

[0015] S21, extracting the light intensity time series from the monitoring group, and constructing different time delays by analyzing the correlation between the light intensities Gq at different times Autocorrelation function of light intensity under , which can be obtained by:

[0016] ;

[0017] In the formula, Represents the total number of light intensity data points; represents the light intensity monitored at time t; Indicates that at time t+ The light intensity monitored at the time Indicates time delay;

[0018] S22, according to the different time delays obtained in S21 Autocorrelation function of light intensity under , get the light intensity autocorrelation function The curve diagram is based on the autocorrelation function of light intensity , find the light intensity autocorrelation function Drop to the maximum The time delay is marked as the light intensity characteristic change time. , based on the light intensity characteristic change time , determine the lower limit of the acquisition frequency caused by the light intensity Gq , the lower limit of the acquisition frequency caused by the light intensity Gq Obtained through the following forms:

[0019] ;

[0020] In the formula, Indicates the time of change of light intensity characteristics;

[0021] S23, extracting the temperature time series from the monitoring group, using fast Fourier transform, decomposing the temperature time series data into a combination of sine and cosine waves of different frequencies, each frequency component has a corresponding amplitude, so as to perform fast Fourier transform (FFT) operation to obtain the spectrum of the temperature time series data , and find the spectrum The frequency level with the highest energy content , based on the frequency with the highest energy share , determine the lower limit of the acquisition frequency caused by temperature Wz , the lower limit of the acquisition frequency caused by the temperature Wz Obtained through the following forms:

[0022] ;

[0023] In the formula, represents pi, Indicates the frequency level with the highest energy share.

[0024] Optionally, the specific steps of S2 also include:

[0025] S24. Obtain historical data according to the data recording device equipped in the photovoltaic system, collect historical fault records of the photovoltaic array according to the historical data, and count the fault types and the frequency of occurrence of each fault type, and select the maximum value of the frequency of occurrence of the fault type as the lower limit of the acquisition frequency caused by the fault occurrence speed .

[0026] Optionally, the specific steps of S2 further include:

[0027] S25. Based on the lower limit of the acquisition frequency caused by the light intensity Gq obtained in S22 - S24 , the lower limit of the acquisition frequency caused by the temperature Wz and the lower limit of the acquisition frequency caused by the fault occurrence speed , use the maximum value method to determine the acquisition frequency Cp. The specific content is: .

[0028] Optionally, the specific steps of S3 include:

[0029] S31. According to the value of the acquisition frequency Cp obtained in S25, perform data acquisition on the photovoltaic array to obtain the current and voltage of the photovoltaic array at different times. After summarization and in the order before and after acquisition, generate the first data point , the second data point , the third data point ,..., the nth data point , and plot the current and voltage of the acquired photovoltaic array into an IV curve. Among them, , , ,..., respectively represent the voltage values at the first data point, the voltage values at the second data point, the voltage values at the third data point,..., the voltage values at the nth data point, , , ,..., respectively represent the current values at the first data point, the current values at the second data point, the current values at the third data point,..., the current values at the nth data point;

[0030] S32. According to the IV curve, use the forward difference method to estimate the first derivative of the ith data point in the IV curve , the first derivative of the ith data point in the IV curve ;

[0031] S33. Based on the first derivatives of each data point calculated in S32 , use the forward difference method again to sequentially estimate the second derivative of the ith data point in the IV curve , the second derivative of the ith data point in the IV curve ;

[0032] S34. According to the content in S33, traverse the sequence of the second derivatives , which are respectively the second derivative of the first data point , the second derivative of the second data point , the second derivative of the third data point , ..., the second-order derivative of the n-1th data point , and based on the second-order derivative of each data point The state change of the local shadow fault is preliminarily marked. The specific content is: if the second-order derivative When it changes from positive to negative, >0 and When <0, it will be between and The positions on the IV curve between the two are preliminarily marked, and after statistics, a preliminary marking set is obtained.

[0033] Optionally, the S3 specific steps also include:

[0034] S35. Based on the degree to which the IV curve tends to be rectangular, obtain the squareness F, wherein the squareness F is obtained by the following formula:

[0035] ;

[0036] In the formula, represents the maximum power point power, represents the open circuit voltage, Indicates short-circuit current.

[0037] Optionally, the S3 specific steps also include:

[0038] S36, using convolutional neural network technology to build a data model, and input the squareness F and the preliminary label set into the data model, and after dimensionless processing, fit and output the determination factor Pz, the determination factor Pz is obtained by the following formula:

[0039] ;

[0040] In the formula, represents the number of labels in the preliminary labeling set, and are weight values, represents the correction constant, and The specific value is set by the user according to the situation;

[0041] S37, presetting a determination threshold K, and comparing the determination threshold K with the determination factor Pz to determine the possibility of a local shadow fault occurring in the photovoltaic array, the specific contents are as follows:

[0042] If the determination factor Pz exceeds the determination threshold K, it is determined that there is a local shadow fault in the photovoltaic array;

[0043] If the determination factor Pz does not exceed the determination threshold K, it is determined that no local shadow fault exists in the photovoltaic array.

[0044] Optionally, the specific steps of S4 include:

[0045] S41, when it is determined that there is a local shadow fault in the photovoltaic array, multiple groups of local maximum power points are determined in combination with the IV curve, and IV characteristic data of each local maximum power point is obtained according to the multiple groups of local maximum power points, the IV characteristic data including the current and voltage of each local maximum power point;

[0046] S42, based on the IV characteristic data of each local maximum power point, combined with the disturbance observation method, dynamically match the output of the photovoltaic array with the input of the inverter.

[0047] Optionally, the specific steps of S5 include:

[0048] S51, extracting from historical data several groups of time periods when the photovoltaic array does not have a local shadow fault and several groups of time periods when the photovoltaic array has a local shadow fault, and based on the IV characteristic data in the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, performing diagnostic test operations on the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, and determining the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array has a local shadow fault that still indicate the presence of a local shadow fault through S36 and the method and comparison content of obtaining the determination factor Pz in S36, and marking it as a true positive rate ZX, and the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array does not have a local shadow fault that still indicate the presence of a local shadow fault, and marking it as a false positive rate JX;

[0049] S52. Based on the true positive rate ZX and the false positive rate JX, obtain the positive likelihood ratio ZJ, which is specifically obtained by the following formula:

[0050] ;

[0051] S53. If the positive likelihood ratio ZJ≥10, verify that the current data model is credible.

[0052] The present invention provides a photovoltaic array fault detection method based on IV characteristics and data model, which has the following beneficial effects:

[0053] (1) This method first determines the acquisition frequency Cp by comprehensively considering dynamic characteristics, historical faults and occurrence speed to ensure that key fault information is not missed. S3 relies on the second-order derivative analysis and squareness calculation of the IV curve, combined with the data model, to accurately identify local shadow faults, changing the previous situation where fault judgment relies on a single indicator and has poor accuracy, and further improving the reliability and timeliness of fault detection. In terms of energy conversion efficiency optimization, S4 determines the local maximum power point based on the judgment factor Pz and the IV curve, and dynamically matches the photovoltaic array output with the inverter input. This enables the photovoltaic system to quickly adjust to a better working state under complex working conditions, such as when some components are affected by shadows, to reduce power loss. Compared with the traditional fixed matching method, it further improves power generation efficiency and reduces energy waste. The model verification mechanism initiated by S5 can verify the accuracy and stability of the model under different scenarios and time spans, which avoids misjudgment or missed judgment caused by model deviation, ensures the long-term stable operation of the photovoltaic system, reduces operation and maintenance costs, and enhances the economic benefits and sustainability of the entire photovoltaic power station.

[0054] (2) S21 constructs the autocorrelation function of light intensity, deeply analyzes the correlation between light intensity at different times, and intuitively presents its correlation. S22 determines the characteristic change time of light intensity based on the autocorrelation function curve, and derives the lower limit of the acquisition frequency based on this time, ensuring that data is collected at least once within the key change period of light intensity, which can effectively capture the dynamic changes of light intensity and prevent key information from being missed. S23 uses fast Fourier transform to decompose complex temperature data into a combination of sine and cosine waves of different frequencies for the temperature time series. By analyzing the frequency with the highest energy share in the spectrum, the lower limit of the acquisition frequency caused by temperature is accurately determined. This method of determining the lower limit of the acquisition frequency based on the characteristics of light and temperature data not only fully considers the impact of environmental factors on the performance of photovoltaic arrays, but also realizes the rational use of resources, improves the effectiveness and pertinence of data acquisition, and provides accurate and efficient data support for subsequent photovoltaic array fault detection, performance analysis and other work.

[0055] (3) S32 and S33 use the forward difference method to cleverly estimate the first-order and second-order derivatives. Forward difference approximates the derivative with the slope of adjacent data points to achieve quantitative analysis of the change trend of the IV curve. The first-order derivative reflects the change in the slope of the curve, and the second-order derivative reveals the rate of change of the slope. These key information are closely related to the working state of the photovoltaic array. Through derivative calculation, the performance change clues implied by the IV curve are deeply explored from a mathematical level. S34 efficiently marks local shadow faults based on the change in the state of the second-order derivative. When the second-order derivative changes from positive to negative, the location where the fault may exist in the IV curve is accurately located. This method uses the principle that local shadow faults change the shape of the IV curve, causing specific changes in the second-order derivative, and accurately delineates the fault area. Compared with traditional methods, it gets rid of the limitations of relying solely on manual experience or simple threshold judgment. It is completely based on mathematical analysis and data processing, which greatly improves the accuracy and timeliness of fault detection. Through this process, potential fault points can be quickly locked, local shadow faults can be handled in a timely manner, and damage to power generation efficiency can be reduced. At the same time, it provides direction for subsequent fault diagnosis and repair, reduces operation and maintenance costs, effectively ensures stable and efficient operation of the photovoltaic system, and improves the overall economic benefits and reliability of the photovoltaic power station.

[0056] (4) In terms of fault feature quantification, S35 converts the shape characteristics of the IV curve into measurable values ​​by calculating the squareness F. When local shadows appear and cause the IV curve to be distorted, the squareness F can accurately reflect this change, making the fault characteristics no longer vague, and providing a clear quantitative basis for subsequent analysis. Compared with traditional qualitative analysis, the accuracy and operability of fault judgment are greatly improved. In data processing and model construction, S36 uses convolutional neural network technology, takes the squareness F and the preliminary label set as input, and enhances the compatibility and comparability of the data through dimensionless processing, further realizing that the dynamic response speed of the inverter can keep up with the changes in the output characteristics of the photovoltaic array. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the present application or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present application.

[0058] Figure 1 The present invention is a schematic flow chart of a photovoltaic array fault detection method based on IV characteristics and data models. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and advantages of this application clearer, the technical solution of this application will be described clearly and completely in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, but not all of them. Based on the embodiments in this application, ordinary technicians in this field can make the following inventions without creative work.

[0060] All other embodiments obtained under the premise of the above-mentioned embodiment belong to the scope of protection of this application.

[0061] Example 1

[0062] See also Figure 1 The present invention provides a photovoltaic array fault detection method based on IV characteristics and data model, comprising the following steps:

[0063] S1. Using several groups of monitoring instruments in advance, real-time monitoring of relevant condition data of the photovoltaic array is performed within a set monitoring period, and a monitoring group is generated after data processing;

[0064] S2. Based on the monitoring group, analyze the impact of the dynamic characteristics of the photovoltaic array on the acquisition frequency, and combine historical data to collect historical fault records of the photovoltaic array, count the frequencies of various faults (such as short circuit, open circuit, local shadow, etc.), analyze the impact of the occurrence speed of each fault type on the acquisition frequency, and determine the acquisition frequency Cp by the maximum value method;

[0065] S3. Based on the acquisition frequency Cp value, data is collected from the photovoltaic array. After the data is collected, the IV curve is drawn. According to the IV curve, the forward difference method is used to estimate the first-order derivative of the discrete data points in the IV curve. and the second-order derivative , based on the second-order derivative The state changes of the local shadow fault are preliminarily marked, and the second-order derivative is traversed When we find that the second-order derivative changes from positive to negative, we can infer that a local maximum may appear near this position, make preliminary statistical marks, and obtain the squareness F based on the degree to which the IV curve tends to be rectangular. Combined with the trained data model, the output judgment factor Pz is fitted to determine the possibility of local shadow failure in the photovoltaic array.

[0066] S4, based on the determination factor Pz and the IV curve, determining multiple groups of local maximum power points to obtain IV characteristic data of each local maximum power point, and dynamically matching the output of the photovoltaic array with the input of the inverter based on the IV characteristic data of each local maximum power point to ensure efficient energy conversion;

[0067] S5. Start the model verification mechanism to verify the credibility of the model.

[0068] In this embodiment, multiple groups of monitoring instruments are used in S1 to monitor the relevant condition data of the photovoltaic array in real time within the set period, and a monitoring group is generated, which provides a multi-faceted and rich data basis for subsequent analysis, which makes the monitoring of the operating status of the photovoltaic array more detailed and can capture more potential fault information. Scientifically set the acquisition frequency: Based on the monitoring group, S2 comprehensively analyzes the dynamic characteristics of the photovoltaic array, historical fault records, and the impact of the speed of occurrence of various faults on the acquisition frequency, and uses the maximum value method to determine the acquisition frequency Cp. This scientific acquisition frequency setting method not only avoids data redundancy and resource waste caused by too high acquisition frequency, but also prevents the omission of key fault information due to too low acquisition frequency. For example, for short-circuit faults with a fast occurrence speed, by accurately setting the acquisition frequency, the electrical parameter changes at the moment of short circuit can be captured in time, providing strong support for rapid fault location and processing. Efficient local shadow fault detection: In S3, the forward difference method is used to perform derivative analysis on the IV curve, combined with the second-order derivative state change and the IV curve squareness F, and then the trained data model is used to fit the output judgment factor Pz, which can efficiently and accurately judge the possibility of local shadow faults. Compared with traditional fault detection methods, this method based on multi-parameter fusion and data model analysis greatly improves the accuracy and timeliness of local shadow fault detection. Dynamic matching of maximum power point: In S4, multiple groups of local maximum power points are determined based on the decision factor Pz and IV curve, and their IV characteristic data are obtained to dynamically match the photovoltaic array output with the inverter input. This process enables the photovoltaic array to always work in a better power output state, ensuring efficient energy conversion. For example, when the photovoltaic array is affected by local shadows, by quickly finding a new local maximum power point and matching the inverter input with it, the power loss caused by the shadow can be minimized and the power generation efficiency of the entire photovoltaic system can be improved. Compared with the system without dynamic matching, the power generation can be significantly improved. Model verification mechanism: The model verification mechanism is started in S5 to verify the credibility of the entire fault detection model. This mechanism ensures the reliability and stability of the model in practical applications.

[0069] Example 2

[0070] Please refer to Figure 1 , specifically: S1 specific steps include:

[0071] S11, deploying several groups of monitoring instruments around the photovoltaic array in advance, and monitoring relevant condition data of the photovoltaic array in real time during the set monitoring period, wherein the relevant condition data includes the light intensity Gq and temperature Wz at each monitoring moment; the several groups of monitoring instruments include photodiode sensors, current sensors, voltage sensors and temperature sensors;

[0072] S12. Check whether there are abnormal values ​​and missing values ​​in the relevant condition data collected in S11, and use statistical methods (such as the 3-sigma principle) to correct the abnormal values, and use interpolation methods (such as spline interpolation) to supplement the missing values. Finally, the processed relevant condition data is used as the monitoring group.

[0073] The specific steps of S2 include:

[0074] S21, extracting the light intensity time series from the monitoring group, and constructing different time delays by analyzing the correlation between the light intensities Gq at different times Autocorrelation function of light intensity under , which can be obtained by:

[0075] ;

[0076] In the formula, Represents the total number of light intensity data points; represents the light intensity monitored at time t; Indicates that at time t+ The light intensity monitored at the time represents the time delay, and its unit is the same as that of the time series t. It reflects the time interval for comparing the light intensity data when calculating the autocorrelation function. For example, when =0, It is actually the variance of the light intensity data (because , at this time, the average value of the square of the light intensity at the same time is calculated, that is, the square of the variance plus the mean, which is the variance when the mean is 0). >0 o'clock, It measures the correlation between light intensities separated in time;

[0077] S22, according to the different time delays obtained in S21 Autocorrelation function of light intensity under , get the light intensity autocorrelation function The curve diagram is based on the autocorrelation function of light intensity , find the light intensity autocorrelation function Drop to the maximum The time delay is marked as the light intensity characteristic change time. , this time scale reflects a "periodic" characteristic of light intensity changes, that is, after After a certain period of time, the change pattern of light intensity has changed significantly compared with the previous one. In order to fully capture the change of light intensity, the acquisition frequency should be high enough to capture the characteristic change time of light intensity. At least one data set is collected within a period of 1 second, which ensures that the main changes in light intensity are not missed. , determine the lower limit of the acquisition frequency caused by the light intensity Gq , the lower limit of the acquisition frequency caused by the light intensity Gq Obtained through the following forms:

[0078] ;

[0079] In the formula, Indicates the time of light intensity characteristic change. The frequency in this formula indicates the minimum acquisition frequency that the data acquisition device should reach in order to adapt to the change of light intensity. For example, if = 10 minutes, then = Times / minute, that is, data is collected at least once every 10 minutes;

[0080] along with increase, The value of will gradually decrease, which reflects that the temporal correlation of light intensity gradually weakens;

[0081] S23, extracting the temperature time series from the monitoring group, using fast Fourier transform, decomposing the temperature time series data into a combination of sine and cosine waves of different frequencies, each frequency component has a corresponding amplitude (energy), so as to perform fast Fourier transform (FFT) operation to obtain the spectrum of the temperature time series data , and find the spectrum The frequency level with the highest energy content , based on the frequency with the highest energy share , determine the lower limit of the acquisition frequency caused by temperature Wz , the lower limit of the acquisition frequency caused by the temperature Wz Obtained through the following forms:

[0082] ;

[0083] In the formula, represents pi, Indicates the frequency level with the highest energy share.

[0084] Spectrum It expresses the amplitude (or energy) of the temperature variation components of different frequencies (in radians per second). For example, if The value of is large, indicating that at a frequency of The component accounts for a large proportion in the temperature change, that is, the temperature change contains more periodic fluctuations of this frequency.

[0085] In the temperature change of photovoltaic array, the fluctuation of different frequencies has different influence on photovoltaic performance. Represents the main periodic or fluctuating characteristics of temperature change. The temperature fluctuation corresponding to this frequency component is relatively significant, and may have a greater impact on the performance of the photovoltaic array (such as output voltage, current, etc.). For example, if The corresponding period is one day (due to the temperature changes between day and night), so this temperature change period will have an important impact on the daily power generation of the photovoltaic array.

[0086] In signal processing, frequency content (in Hz) and angular frequency (unit is radians per second) is So, to convert angular frequency to frequency, we need to divide by , this frequency is the frequency corresponding to the main frequency component of temperature change. It means that in order to capture the main dynamic characteristics of temperature change, the acquisition frequency should at least reach this value. If the acquisition frequency is lower than , the most important fluctuations in temperature changes may be missed, thus affecting the monitoring and analysis of changes in photovoltaic array performance.

[0087] If the main frequency components are not considered and the acquisition frequency is set arbitrarily, two situations may occur: first, the acquisition frequency is too high, which will collect a large amount of unnecessary data and increase storage and processing costs; second, the acquisition frequency is too low, which will miss key information about temperature changes and fail to accurately assess the impact of temperature changes on the photovoltaic array. By determining the lower limit of the acquisition frequency corresponding to the frequency component with the highest energy share, we can obtain enough information to reflect the important impact of temperature on the photovoltaic array without excessive data collection.

[0088] In this embodiment, in terms of data collection, S11 deploys multiple sets of monitoring instruments around the photovoltaic array to monitor relevant condition data such as light intensity and temperature in real time, thereby ensuring the comprehensiveness and real-time nature of the data source. In S12, the collected data is corrected for outliers and supplemented for missing values, and the 3-sigma principle and spline interpolation method are used to effectively improve the data quality, provide a reliable basis for subsequent precise analysis, and avoid analytical deviations caused by data errors or missing data. The process of determining the lower limit of the acquisition frequency has significant advantages. S21 uses the light intensity time series to construct an autocorrelation function, clearly presenting the degree of correlation between light intensity under different time delays. By finding the time delay when the autocorrelation function of light intensity drops to the maximum value 1 / e, marked as the characteristic change time of light intensity, it provides a key basis for determining the lower limit of acquisition frequency. For temperature time series, S23 uses fast Fourier transform to decompose the data into a combination of sine and cosine waves of different frequencies, and finds the frequency with the highest energy proportion in the spectrum, so as to determine the lower limit of acquisition frequency caused by temperature. This method of determining the lower limit of acquisition frequency based on data characteristics fully considers the impact of light intensity and temperature on the performance of photovoltaic arrays, ensuring that the acquisition frequency can meet the needs of capturing changes in environmental parameters and avoid unnecessary over-acquisition. While improving the effectiveness of data acquisition, it also improves the system operation efficiency and resource utilization, providing strong support for the accuracy and timeliness of photovoltaic array fault detection.

[0089] Example 3

[0090] Please refer to Figure 1 Specifically: S2 includes the following specific steps:

[0091] S24. Obtain historical data based on the data recording equipment equipped in the photovoltaic system, wherein the data recording equipment includes but is not limited to inverters, data collectors and smart meters, etc. The data collector refers to a device specifically used to collect various sensor data, which can be connected to multiple different types of sensors, such as light intensity sensors, temperature sensors, irradiation sensors, etc. The data collector collects data from the sensors at a preset time interval (collection frequency) and performs preliminary processing on the data, such as analog-to-digital conversion (if it is an analog signal sensor), data format conversion, etc.; based on the historical data, collect historical fault records of the photovoltaic array, and count the fault types (such as short circuit, open circuit, partial shadow, etc.), as well as the frequency of occurrence of each fault type, and select the maximum value of the fault type frequency as the lower limit of the collection frequency caused by the fault occurrence speed. .

[0092] The specific steps of S2 also include:

[0093] S25, based on the lower limit of the acquisition frequency caused by the light intensity Gq obtained in S22-S24 , the lower limit of the acquisition frequency caused by temperature Wz The lower limit of the acquisition frequency caused by the speed of the fault , using the maximum value method, determine the acquisition frequency Cp, the specific content is:

[0094] The idea behind this approach is to ensure that the acquisition frequency can meet relatively demanding requirements. For example, if the light intensity changes very quickly, the corresponding acquisition frequency lower limit is very high, then collecting at this highest frequency can ensure that enough data can be obtained even when the light intensity changes rapidly. At the same time, this frequency is also sufficient to cope with temperature changes and fault detection, because it has met relatively stringent acquisition requirements.

[0095] In this embodiment, in the process of acquiring historical data and analyzing faults, S24 integrates the data records of devices such as inverters, data collectors, and smart meters, collects historical data extensively, and comprehensively acquires the historical fault records of the photovoltaic array. By counting the occurrence frequency of each fault type (short circuit, open circuit, partial shadow, etc.), and taking the highest frequency as the lower limit of the acquisition frequency caused by the fault occurrence speed, it accurately grasps the key requirements of fault detection, ensuring that during the high-incidence period of faults, data acquisition can timely capture the subtle changes that may cause faults, and provide data support for early warning and rapid diagnosis of faults. In determining the final acquisition frequency, S25 uses the maximum value method to comprehensively consider the lower limit of the acquisition frequency caused by light intensity, temperature, and fault occurrence speed. The light intensity changes rapidly, and its lower limit determines the minimum frequency requirement for capturing light dynamics; the lower limit of temperature reflects the acquisition accuracy required for the impact of temperature on the performance of the photovoltaic array; and the lower limit of faults guarantees the amount of data collected during the high-incidence period of faults. The maximum value method is used to determine the acquisition frequency Cp with the goal of meeting relatively stringent requirements, achieving multiple goals at one stroke. On the one hand, it can ensure that the data acquisition system will not miss any key changes when the light intensity changes rapidly, providing detailed data for studying the impact of light on the performance of photovoltaic arrays. On the other hand, since this frequency is higher than the minimum frequency required for temperature changes and fault detection, it can also provide sufficient and timely data when dealing with temperature fluctuations and fault detection scenarios. This not only optimizes data collection efficiency and avoids information omissions due to insufficient collection, but also prevents waste of resources due to excessive collection, further improving the overall performance of photovoltaic array monitoring and fault detection.

[0096] Example 4

[0097] Please refer to Figure 1 , specifically: S3 specific steps include:

[0098] S31, according to the acquisition frequency Cp value obtained in S25, data is collected on the photovoltaic array to obtain the current and voltage of the photovoltaic array at different times, and the first data points are generated in sequence after being summarized and in accordance with the order before and after the acquisition. , the second data point , the third data point , ..., the nth data point , and draw the collected current and voltage of the photovoltaic array into an IV curve, where , , ,..., represent the voltage value at the first data point, the voltage value at the second data point, the voltage value at the third data point, ..., the voltage value at the nth data point, respectively. , , ,..., Respectively represent the current value at the first data point, the current value at the second data point, the current value at the third data point, ..., the current value at the nth data point;

[0099] S32, according to the IV curve, using the forward difference method, estimate the first-order derivative of the i-th data point in the IV curve , the first derivative of the ith data point in the IV curve , this is because the basic idea of ​​forward difference is to use the slope between two adjacent data points to approximate the derivative of the point;

[0100] S33, based on the first-order derivative of each data point calculated in S32 , use the forward difference method again to estimate the second-order derivative of the i-th data point in the IV curve , the second-order derivative of the ith data point in the IV curve , where the denominator This is because the second-order derivative involves the difference between two adjacent first-order derivatives, and the voltage interval corresponding to these two first-order derivatives is from arrive .

[0101] S34. According to the content in S33, traverse the second-order derivative The sequence is the second-order derivative of the first data point. , the second derivative of the second data point , the second derivative of the third data point , ..., the second-order derivative of the n-1th data point , and based on the second-order derivative of each data point The state change of the local shadow fault is preliminarily marked. The specific content is: if the second-order derivative When it changes from positive to negative, >0 and When <0, it will be between and The positions on the IV curve between the two are preliminarily marked, and after statistics, a preliminary marking set is obtained.

[0102] In this embodiment, during the data collection and curve drawing stage, S31 collects the current and voltage of the photovoltaic array according to the precise collection frequency Cp, generates a series of ordered data points, and draws the IV curve accordingly. This process ensures the integrity and accuracy of the data. The IV curve can accurately reflect the electrical performance state of the photovoltaic array at different times, providing an intuitive and reliable basis for subsequent in-depth analysis. The steps of calculating the derivative using the forward difference method (S32 and S33) are highly innovative and practical. Through the forward difference method, the first-order derivative of each data point in the IV curve is first estimated, and then the second-order derivative is calculated based on the first-order derivative. This method cleverly uses the relationship between adjacent data points to discretize the continuous IV curve and effectively extract the curve change trend information. The first-order derivative reflects the slope change of the IV curve, and the second-order derivative further reveals the rate of slope change, and these changes are closely related to the working state of the photovoltaic array. In the fault detection link, S34 traverses the second-order derivative sequence and preliminarily marks the local shadow fault according to the characteristics of the second-order derivative changing from positive to negative. This method uses the principle that the local shadow fault causes the IV curve to deform, which in turn causes a specific change in the second-order derivative. By accurately marking the locations where local shadow faults may exist and forming a preliminary marking set, the accuracy and pertinence of fault detection are further improved. Compared with traditional fault detection methods, it no longer relies on manual experience or simple threshold judgments, but is based on scientific mathematical analysis and data processing, which can locate potential fault areas more quickly and accurately. This not only helps to timely discover and deal with local shadow faults and reduce their impact on the power generation efficiency of photovoltaic arrays, but also provides a clear direction for subsequent fault diagnosis and repair work, effectively reducing operation and maintenance costs and improving the reliability and stability of photovoltaic systems.

[0103] Example 5

[0104] Please refer to Figure 1 , specifically: S3 specific steps also include:

[0105] S35. Based on the degree to which the IV curve tends to be rectangular, obtain the squareness F, wherein the squareness F is obtained by the following formula:

[0106] ;

[0107] In the formula, represents the maximum power point power, represents the open circuit voltage, Indicates short-circuit current.

[0108] In physical terms, the squareness F reflects the extent to which the photovoltaic array can approach the ideal rectangular IV curve (ideally, when the IV curve is rectangular, the squareness F (i.e., the fill factor) is 1, indicating that the photovoltaic array can output all light energy at maximum power).

[0109] When there is a local shadow fault, the IV curve will have multiple local maximum power points (LMPP). In this case, each LMPP has a corresponding power, which is calculated by multiplying the voltage and current at that point. When calculating the squareness F, it is usually based on the normal maximum power point power when there is no local shadow fault (if the power in the normal state is already known), or the maximum power point power that is most likely to represent the performance of the entire array determined after an overall analysis of the current IV curve is calculated and compared.

[0110] When a local shadow fault occurs in a photovoltaic array, some photovoltaic cells are shaded and their output characteristics change. The photocurrent generated by the shaded cells decreases, which is equivalent to forming a "current bottleneck" in the circuit of the entire photovoltaic array. In this case, the IV curve of the entire photovoltaic array will be distorted. Due to the mismatch of current-voltage characteristics between different cells caused by local shadows, the maximum power point power decreases, and the open circuit voltage and short circuit current will also be affected to a certain extent (the open circuit voltage may decrease slightly, and the short circuit current decreases due to the reduced contribution of the shaded part).

[0111] Open circuit voltage refers to the output voltage of the photovoltaic array when there is no external load (i.e. the current is equal to 0). From a physical point of view, it is the potential difference formed by the accumulation of photogenerated carriers inside the photovoltaic cell at both ends of the cell when there is no loop current. The size of this potential difference is related to factors such as the bandgap width of the photovoltaic material, light intensity, and temperature.

[0112] Short-circuit current refers to the output current of a photovoltaic array when the output end is short-circuited (i.e., the voltage is equal to 0). At this time, the photogenerated carriers almost all form current under the action of the internal electric field, and there is no potential difference to hinder the movement of the carriers. The short-circuit current mainly depends on the light intensity and the area of ​​the photovoltaic array (under the same material and other conditions, the larger the area, the more photons are absorbed, and the greater the short-circuit current).

[0113] The specific steps of S3 also include:

[0114] S36, using convolutional neural network technology to build a data model, and input the squareness F and the preliminary label set into the data model, and after dimensionless processing, fit and output the determination factor Pz, the determination factor Pz is obtained by the following formula:

[0115] ;

[0116] In the formula, represents the number of labels in the preliminary labeling set, and are weight values, Represents the correction constant, where 0< <1,0< <1, and The specific value is set by the user according to the situation;

[0117] S37, presetting a determination threshold K, and comparing the determination threshold K with the determination factor Pz to determine the possibility of a local shadow fault occurring in the photovoltaic array, the specific contents are as follows:

[0118] If the determination factor Pz exceeds the determination threshold K, it is determined that there is a local shadow fault in the photovoltaic array;

[0119] If the determination factor Pz does not exceed the determination threshold K, it is determined that no local shadow fault exists in the photovoltaic array.

[0120] In this embodiment, S35 intuitively reflects the degree of proximity between the IV curve of the photovoltaic array and the ideal rectangle by calculating the squareness F. When a local shadow fault occurs, the output characteristics of the photovoltaic array change, the IV curve is distorted, and the change in squareness F sensitively captures these anomalies. Compared with the traditional method of judging faults based on experience or a single indicator, using squareness as a reference can more accurately quantify array performance and provide strong data support for fault detection. S36 uses convolutional neural network technology to construct a data model and combines squareness F with a preliminary labeling set. Dimensionless processing ensures that data is analyzed at a uniform scale and improves model accuracy. By fitting the output judgment factor Pz through the model, the influence of various factors on the fault is comprehensively considered, which greatly enhances the ability to identify local shadow faults. S37 sets the judgment threshold K and compares it with the judgment factor Pz to simply and intuitively judge the possibility of the fault. This method is clear and provides operators with a quick and effective decision-making basis. When the determination factor Pz exceeds the threshold value K, a fault warning can be issued in time, allowing the operation and maintenance personnel to respond quickly and reduce the power generation loss caused by local shadow faults; if it does not exceed the threshold, you can rest assured that the photovoltaic array is not affected by this fault, avoiding unnecessary inspections and maintenance, effectively improving operation and maintenance efficiency and reducing operation and maintenance costs.

[0121] Example 6

[0122] Please refer to Figure 1 , specifically: S4 specific steps include:

[0123] S41, when it is determined that there is a local shadow fault in the photovoltaic array, multiple groups of local maximum power points are determined in combination with the IV curve, and IV characteristic data of each local maximum power point is obtained according to the multiple groups of local maximum power points, the IV characteristic data including the current and voltage of each local maximum power point;

[0124] S42. Based on the IV characteristic data of each local maximum power point, combined with the perturbation observation method, the output of the photovoltaic array is dynamically matched with the input of the inverter to ensure efficient energy conversion.

[0125] The perturbation observation method is a common MPPT algorithm. When there are multiple LMPPs, the inverter will perform small periodic perturbations on the working point of the photovoltaic array. For example, it will first slightly increase the input voltage and observe the change in power. If the power increases, it means that the perturbation direction is correct, and continue to adjust in this direction; if the power decreases, change the perturbation direction. Through this continuous trial and adjustment, the algorithm can find the point with the highest power or a relatively optimal working point among multiple LMPPs.

[0126] The specific steps of S5 include:

[0127] S51, extracting from historical data several groups of time periods when the photovoltaic array does not have a local shadow fault and several groups of time periods when the photovoltaic array has a local shadow fault, and based on the IV characteristic data in the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, performing diagnostic test operations on the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, and determining the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array has a local shadow fault that still indicate the presence of a local shadow fault through S36 and the method and comparison content of obtaining the determination factor Pz in S36, and marking it as a true positive rate ZX, and the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array does not have a local shadow fault that still indicate the presence of a local shadow fault, and marking it as a false positive rate JX;

[0128] S52. Based on the true positive rate ZX and the false positive rate JX, obtain the positive likelihood ratio ZJ, which is specifically obtained by the following formula:

[0129] ;

[0130] S53. If the positive likelihood ratio ZJ≥10, verify that the current data model is credible.

[0131] In this embodiment, in terms of energy conversion, after determining that there is a local shadow fault, S41 accurately locates multiple groups of local maximum power points and obtains their IV characteristic data, providing key information for system adjustment. S42 combines the perturbation observation method to dynamically match the inverter with the photovoltaic array output. Faced with multiple local maximum power points, the inverter continuously tests the power change through small periodic disturbances, and then finds a relatively optimal working point. This can tap the power generation potential of the photovoltaic array as much as possible under the influence of local shadow faults, ensure efficient energy conversion, greatly improve power generation efficiency, reduce power losses caused by shadows, and allow the photovoltaic system to operate stably under complex working conditions. At the model verification level, S51 extracts different state periods from historical data, uses the method of S36 to perform diagnostic tests, and calculates the true positive rate ZX and the false positive rate JX respectively. This process is rigorous and comprehensive, and a large amount of actual data is used to verify the judgment accuracy of the model. S52 obtains the positive likelihood ratio ZJ based on ZX and JX, which intuitively reflects the performance of the model. When ZJ ≥ 10 in S53, it fully proves that the current data model has a high degree of credibility, which means that the model can reliably detect local shadow faults in practical applications and reduce the risk of misjudgment and missed judgment. Operation and maintenance personnel can make correct decisions based on this model to avoid blind maintenance or ignoring faults, thereby saving time and labor costs and improving the overall reliability and economic benefits of the photovoltaic system.

[0132] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A photovoltaic array fault detection method based on IV characteristics and data model, characterized in that: The following steps are involved: S1. Using several groups of monitoring instruments in advance, real-time monitoring of relevant condition data of the photovoltaic array is performed within a set monitoring period, and a monitoring group is generated after data processing; S2. Based on the monitoring group, analyze the impact of the dynamic characteristics of the photovoltaic array on the acquisition frequency, and analyze the impact of the occurrence speed of each fault type on the acquisition frequency in combination with historical data, and determine the acquisition frequency Cp by the maximum value method; S3. Based on the acquisition frequency Cp value, data is collected from the photovoltaic array. After the data is collected, the IV curve is drawn. According to the IV curve, the forward difference method is used to estimate the second-order derivative of the discrete data points in the IV curve. , based on the second-order derivative The state changes of the local shadow fault are preliminarily marked, and the second-order derivative is traversed The sequence is statistically preliminarily marked, and the squareness F is obtained based on the degree to which the IV curve tends to be rectangular. Combined with the trained data model, the output judgment factor Pz is fitted to determine the possibility of local shadow failure in the photovoltaic array; S4, based on the determination factor Pz and the IV curve, determining multiple groups of local maximum power points to obtain IV characteristic data of each local maximum power point, and dynamically matching the output of the photovoltaic array with the input of the inverter based on the IV characteristic data of each local maximum power point; S5. Start the model verification mechanism to verify the credibility of the model.

2. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 1, characterized in that: The specific steps of S1 include: S11, deploying several groups of monitoring instruments around the photovoltaic array in advance, and monitoring relevant condition data of the photovoltaic array in real time during the set monitoring period, wherein the relevant condition data includes the light intensity Gq and temperature Wz at each monitoring moment; the several groups of monitoring instruments include photodiode sensors, current sensors, voltage sensors and temperature sensors; S12. Check whether the relevant condition data collected in S11 have abnormal values ​​and missing values, and use statistical methods to correct the abnormal values ​​and use interpolation methods to supplement the missing values. Finally, the processed relevant condition data is used as the monitoring group.

3. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 2, characterized in that: The specific steps of S2 include: S21, extracting the light intensity time series from the monitoring group, and constructing different time delays by analyzing the correlation between the light intensities Gq at different times Autocorrelation function of light intensity under , which can be obtained by: ； In the formula, Represents the total number of light intensity data points; represents the light intensity monitored at time t; Indicates that at time t+ The light intensity monitored at the time Indicates time delay; S22, according to the different time delays obtained in S21 Autocorrelation function of light intensity under , get the light intensity autocorrelation function The curve diagram is based on the autocorrelation function of light intensity , find the light intensity autocorrelation function Drop to the maximum The time delay is marked as the light intensity characteristic change time. , based on the light intensity characteristic change time , determine the lower limit of the acquisition frequency caused by the light intensity Gq , the lower limit of the acquisition frequency caused by the light intensity Gq Obtained through the following forms: ； In the formula, Indicates the time of change of light intensity characteristics; S23, extracting the temperature time series from the monitoring group, using fast Fourier transform, decomposing the temperature time series data into a combination of sine and cosine waves of different frequencies, each frequency component has a corresponding amplitude, so as to perform fast Fourier transform (FFT) operation to obtain the spectrum of the temperature time series data , and find the spectrum The frequency level with the highest energy content , based on the frequency with the highest energy share , determine the lower limit of the acquisition frequency caused by temperature Wz , the lower limit of the acquisition frequency caused by the temperature Wz Obtained through the following forms: ； In the formula, represents pi, Indicates the frequency level with the highest energy share.

4. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 3, characterized in that: The specific steps of S2 also include: S24. Obtain historical data according to the data recording device equipped in the photovoltaic system, collect historical fault records of the photovoltaic array according to the historical data, and count the fault types and the frequency of occurrence of each fault type, and select the maximum value of the frequency of occurrence of the fault type as the lower limit of the acquisition frequency caused by the fault occurrence speed .

5. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 4, characterized in that: The specific steps of S2 also include: S25, based on the lower limit of the acquisition frequency caused by the light intensity Gq obtained in S22-S24 , the lower limit of the acquisition frequency caused by temperature Wz The lower limit of the acquisition frequency caused by the speed of the fault , using the maximum value method, determine the acquisition frequency Cp, the specific content is: .

6. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 5, characterized in that: The specific steps of S3 include: S31, according to the acquisition frequency Cp value obtained in S25, data is collected on the photovoltaic array to obtain the current and voltage of the photovoltaic array at different times, and the first data points are generated in sequence after being summarized and in accordance with the order before and after the acquisition. , the second data point , the third data point , ..., the nth data point , and draw the collected current and voltage of the photovoltaic array into an IV curve, where , , ,..., represent the voltage value at the first data point, the voltage value at the second data point, the voltage value at the third data point, ..., the voltage value at the nth data point, respectively. , , ,..., Respectively represent the current value at the first data point, the current value at the second data point, the current value at the third data point, ..., the current value at the nth data point; S32, according to the IV curve, using the forward difference method, estimate the first-order derivative of the i-th data point in the IV curve , the first derivative of the ith data point in the IV curve ; S33, based on the first-order derivative of each data point calculated in S32 , use the forward difference method again to estimate the second-order derivative of the i-th data point in the IV curve , the second-order derivative of the ith data point in the IV curve ; S34. According to the content in S33, traverse the second-order derivative The sequence is the second-order derivative of the first data point. , the second derivative of the second data point , the second derivative of the third data point , ..., the second-order derivative of the n-1th data point , and based on the second-order derivative of each data point The state change of the local shadow fault is preliminarily marked. The specific content is: if the second-order derivative When it changes from positive to negative, >0 and When <0, it will be between and The positions on the IV curve between the two are preliminarily marked, and after statistics, a preliminary marking set is obtained.

7. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 6, characterized in that: The specific steps of S3 also include: S35. Based on the degree to which the IV curve tends to be rectangular, obtain the squareness F, wherein the squareness F is obtained by the following formula: ； In the formula, represents the maximum power point power, represents the open circuit voltage, Indicates short-circuit current.

8. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 7, characterized in that: The specific steps of S3 also include: S36, using convolutional neural network technology to build a data model, and input the squareness F and the preliminary label set into the data model, and after dimensionless processing, fit and output the determination factor Pz, the determination factor Pz is obtained by the following formula: ； In the formula, represents the number of labels in the preliminary labeling set, and are weight values, represents the correction constant, and The specific value is set by the user according to the situation; S37, presetting a determination threshold K, and comparing the determination threshold K with the determination factor Pz to determine the possibility of a local shadow fault occurring in the photovoltaic array, the specific contents are as follows: If the determination factor Pz exceeds the determination threshold K, it is determined that there is a local shadow fault in the photovoltaic array; If the determination factor Pz does not exceed the determination threshold K, it is determined that no local shadow fault exists in the photovoltaic array.

9. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 8, characterized in that: The specific steps of S4 include: S41, when it is determined that there is a local shadow fault in the photovoltaic array, multiple groups of local maximum power points are determined in combination with the IV curve, and IV characteristic data of each local maximum power point is obtained according to the multiple groups of local maximum power points, the IV characteristic data including the current and voltage of each local maximum power point; S42, based on the IV characteristic data of each local maximum power point, combined with the disturbance observation method, dynamically match the output of the photovoltaic array with the input of the inverter.

10. A photovoltaic array fault detection method based on IV characteristics and data model according to claim 9, characterized in that: The specific steps of S5 include: S51, extracting from historical data several groups of time periods when the photovoltaic array does not have a local shadow fault and several groups of time periods when the photovoltaic array has a local shadow fault, and based on the IV characteristic data in the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, performing diagnostic test operations on the several groups of time periods when the photovoltaic array does not have a local shadow fault and the several groups of time periods when the photovoltaic array has a local shadow fault, and determining the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array has a local shadow fault that still indicate the presence of a local shadow fault through S36 and the method and comparison content of obtaining the determination factor Pz in S36, and marking it as a true positive rate ZX, and the proportion of the diagnostic test results of the several groups of time periods when the photovoltaic array does not have a local shadow fault that still indicate the presence of a local shadow fault, and marking it as a false positive rate JX; S52. Based on the true positive rate ZX and the false positive rate JX, obtain the positive likelihood ratio ZJ, which is specifically obtained by the following formula: ； S53. If the positive likelihood ratio ZJ≥10, verify that the current data model is credible.

Images